Chain-linking methods used within the UK

National Accounts

Contact: Release date: Next release:

Steve Drew, Phil Lewis, and 10 October 2016 To be announced
Craig McLaren
philip.a.lewis@ons.gsi.gov.uk

Table of contents
1. Introduction

2. Unchaining

3. Deriving weights for aggregation

4. Chain-linking

5. Tail of series up to Blue Book 2016

6. Re-referencing and the tail calculation from Blue Book 2017

7. National Accounts Identities and consistency checks

8. Summary

9. Acknowledgements

10. Appendix ONS National Accounts notation and terminology

11. References and further reading

Page 1 of 22
1 . Introduction

We produce the UK National Accounts estimates which include the calculation of volume measure series for
economic data.

This is a technical reference paper which describes unchaining and chain-linking methodology used in the
production of volume measures within the UK. Section 10 describes all terminology used in this paper.

A volume measure is a series of economic data from successive years expressed in real terms by calculating the
production volume for each year in the prices of a reference year. The resultant time-series of production figures
has the effects of price changes removed. That is, the effects of monetary inflation or deflation have been
removed so that the series reflects only production volume.

A current price (CP) estimate records the actual or estimated monetary value for a defined period. The current
price estimate is the value expressed in terms of the prices of that period. A time series of CP estimates can be
constructed.

If a suitable deflator (or equivalently referred to as a price index) exists, then applying the deflator to a CP series
the associated constant price (KP) or sometimes referred to as the volume measure (VM) series can be
calculated. The relationship is

value = volume x price

So in this notation

CP = KP x deflator

and then

CP/deflator = KP

Chain volume measure (CVM) estimates are VM estimates obtained by chain-linking. They are the result of
joining together two indices that overlap in one period by rescaling one of them to make its value equal to that of
the other in the same period, thus combining them into single consistent chain volume measure time series.

Corresponding CP and KP series are unchained to produce the previous year’s prices series (PYP) and current
year’s prices series (CYP). The PYP series may be added or subtracted, weighted together or proportioned out to
produce a new PYP series. Similarly the CYP series may be added or subtracted, weighted together or
proportioned out. The resultant PYP series with its corresponding CYP series may then be chain-linked to
produce a CVM estimate.

Eurostat (2013a, Section 2.2.3) defines the base and reference year as:

“– the base year is the year whose current price values are used to weight the price and volume measures
derived at the elementary level of aggregation;

– the reference year is the year which is used for the presentation of a time series of volume data. In a series of
index numbers it is the year that takes the value 100.”

Page 2 of 22
It is important to note that the reference year bears no fixed relation to the years from which weights have been
used and re-referencing is simply the application of a scaling factor to a time series, and does not change the
growth rates in the time series.

The expression "chained volume measure" or “CVM” is used in the UK National Accounts publications to describe
volume measures derived by chain-linking in either index form (that is, set to be 100 for reference year) or in
£million form (that is referenced to the current monetary value in the reference year). This is also sometimes
referred to as a “chain-linked volume” (CLV) or “chained volume” (see Eurostat 2013b, Chapter 6). The “CVM”
notation is used in this article.

For constant price estimates the base period and the reference period coincide. Importantly for chained volume
measures there is only one reference period, but there are many base periods.

The PYP (or CYP) series can be treated in the same way as a fixed base series, as you essentially have a series
of annual fixed base series. A fixed base series uses fixed weights from the defined base period for all
component series. So in this format PYP may be added or subtracted, weighted together or proportioned out.
Chain-linking the PYP series together (using the corresponding CYP series) converts the unchained series which
are expressed in the previous year’s prices into a continuous series expressed in the prices of the reference year.
The weights of a chained volume measure are updated each year and are based on the previous year’s data.
That is, it is not a fixed-weighted index with a single base year. Effectively the chained volume measure has
multiple base periods and one reference period.

Additivity over aggregations is lost when these calculations are made for the chained periods. That is component
series do not necessarily sum to totals. Eurostat (2013a, Section 2.2.3) also covers this point and states: “To
keep all year-to-year growth rates of each variable unchanged when the reference year is changed, one should
re-reference each variable separately, be it an elementary index, a sub-total or an overall aggregate such as
GDP. The consequence is that, in the chained volume data of a fixed reference year, discrepancies will arise
between individual elements and their totals. This is the 'non-additivity' problem”.

In the UK national accounts the reference period used is the last fully balanced year. The choice of reference
period in the UK national accounts is discussed in Documentation on Current Methods used for National
Accounts, ONS (2008):

“The same reference year is used for all national accounts aggregates. When annual chain-linking was introduced
to the UK National Accounts, the decision was made, after consultation with external users, to make the
reference year the same as the last year from which weights have been used in the aggregate time series (and
adopt this reference year with lower level-time series, regardless of the last year of weights which had been used
in their construction).”

Every year we update the UK National Accounts through a process known as annual supply and use balancing.
This brings together detailed data on the three approaches to measuring gross domestic product (GDP). It
balances supply and demand by product, inputs and outputs by industry. Volume series are updated so their
reference and base years are moved forward, usually (but not always) by one year. At the same time, major
methodological or classification changes, may be implemented within the accounts. The publication incorporating
these changes is known as the Blue Book, and the Quarterly National Accounts publication published a month
prior is consistent with Blue Book. The Blue Book also includes changes made to incorporate new data sources,
including those used for annual benchmarking purposes.

Weights to aggregate series are created for all years that have been through at least two annual rounds of supply
use balancing. After the reference year the weights are not yet available. Therefore for years after the reference
year the weights from the reference year are used. Consequently after the reference year the aggregation
becomes a fixed base aggregation, as the final year for which weights exist becomes the base year for all years
thereafter. This is sometimes referred to as a “fixed base tail”. See section 5 for discussion on the treatment of
series tails. For Blue Book 2016, a full set of weights were available for years up and including 2013. That is,
2013 was set as the reference period for Blue Book 2016. (Blue Book 2017 is discussed in section 6).

Page 3 of 22
2 . Unchaining

Unchaining converts a constant price series, for example constant price (KP), volume measure (VM) or chained
volume measure (CVM), with a single reference year into an unchained series where each year is expressed in
the previous year's prices. In our context the “last base year” is the last year which has been balanced through
input-output supply and use tables (see Mahajan, 2006), where all values after that year are expressed in that
year's prices. The last base year is used as the reference period and in the reference period the annual current
price (CP) value = the annual KP value.

Within our processing system, unchaining is achieved using a standard statistical function, UNCHAIN. The inputs
are a KP series and a corresponding CP series. The last base year for these series is stated as a parameter in
the function call as well as whether the output should be a previous year’s prices (PYP) or a current year’s prices
(CYP) series.

The steps of the process required before unchaining are:

collect and derive a time series of CP estimates

calculate a KP equivalent time series, for example, by calculating deflators and then deflating the CP
estimates

for a single KP time series the quarterly values in each year may be added together to generate the annual
value. Similarly the monthly values in each year may be added to generate the quarterly values or the
annual value

calculate annual CP values (that is, sum quarterly input (or monthly) to annual)

calculate annual KP values (that is, sum quarterly input (or monthly) to annual)

calculate PYPs as described below (that is, unchaining)

calculate CYPs as described below (that is, unchaining)

Within our processing systems the UNCHAIN function can also be used for intermediate calculations in order to
derive the required CVM outputs. The outputs from the UNCHAIN function are CYP and PYP estimates which are
consistent so that when chain-linked back again give the input CVM estimate (before benchmarking and
smoothing are applied). Note, in this context for national accounts, that chain-linking is not the opposite of
unchaining because chain-linking includes additional statistical processing steps for benchmarking and smoothing.

The PYP (and CYP) time series may be used in the same way as a fixed base series. Each year of the PYP
series is an annual fixed base series. So that the PYP series may be added or subtracted as well as weighted
together or proportioned out as required to the desired series. Similarly this applies to the CYP series.

2.1 Calculating previous year's prices: for years up to and including the last
base year

The PYP estimates are calculated from the value of the constant price series for that period divided by the value
of the annual constant price series for the year preceding the time period and multiplied by the value of the
annual current price series for that year.

To generate the PYP series for “year y” the annual CP and annual KP values for year y-1 are required. Therefore
the PYP series starts for the year after the first year of the input CP and KP data. This is sometimes referred to
as “losing the first year”.

Page 4 of 22
Note that in the following notation, data for months, quarters or years used are complete full data for that time
period. A separate discussion and treatment is needed for the tail of the time series (see Section 2.3 in this
paper).

Note that for ease of reading the notation PYP(y) is used to represent the PYP value for year y. This is often
written as PYP y.

2.1.1 Notation for unchaining an annual KP series to a PYP series for a given year

where

y-1 is the year previous to year y

PYP(y) is the PYP value for year y

KP(y) is the KP value year y

CP(y-1) is the annual CP value for year y-1

KP(y-1) is the annual KP value for year y-1

2.1.2 Notation for unchaining a quarterly KP series to a PYP series for a given quarter

where

y-1 is the year previous to year y

PYP(q, y) is the PYP value for quarter q, year y

KP(q, y) is the KP value for quarter q, year y

CP(y-1) is the annual CP value for year y-1

KP(y-1) is the annual KP value for year y-1

2.13 Notation for unchaining a monthly KP series to a PYP series for a given month

where

Page 5 of 22
 y-1 is the year previous to year y

PYP(m, y) is the PYP value for month m, year y

KP(m, y) is the KP value for month m, year y

CP(y-1) is the annual CP value for year y-1

KP(y-1) is the annual KP value for year y-1

2.2 Calculating current year's prices: for years up to and including the last
base year

The current year’s price series (CYP) is calculated as the value of the constant price series for that period divided
by the value of the annual constant price series for the year containing the time period and multiplied by the value
of the annual current price series for that year.

2.2.1 Notation for unchaining an annual KP series to a CYP series for a given year

where

CYP(y) is the CYP value for year y

KP(y) is the KP value year y

CP(y) is the annual CP value for year y

2.2.2 Notation for unchaining a quarterly KP series to a CYP series for a given quarter

where

CYP(q, y) is the CYP value for quarter q, year y

KP(q, y) is the KP value for quarter q, year y

CP(y) is the annual CP value for year y

2.2.3 Notation for unchaining a monthly KP series to a CYP series for a given month

where

CYP(m, y) is the CYP value for month m, year y

KP(m, y) is the KP value for month m, year y

CP(y) is the annual CP value for year y

Page 6 of 22
2.3 Calculating previous year's prices and current year's prices for periods
after the last base year (the tail)

For our outputs up to Blue Book 2016 after the last base year, both the PYP and CYP series are defined to be
the value of the constant price series for that period divided by the value of the annual constant price series for
the last base year and multiplied by the value of the annual current price series for the last base year. However,
by definition of the last base year, the annual constant price series equals the value of the annual current price
series for the last base year. Periods after the last base year are often referred to as the tail of the series.
Therefore periods in the tail for outputs up to Blue Book 2016, were not truly a “previous year’s prices” series
despite being labelled as one.

For Blue Book 2016, the last base year will be set to be 2013. See section 5 for further discussion on the
treatment of series tails.

2.3.1 Notation for CYP series and PYP Series from annual KP for periods after the last base
year for a given year

where

PYP(y) is the PYP value for year y

CYP(y) is the CYP value for year y

KP(y) is the KP value year y

2.3.2 Notation for CYP series and PYP Series from quarterly KP for periods after the last
base year for a given quarter

where

PYP(q, y) is the PYP value for quarter q, year y

CYP(q, y) is the CYP value for quarter q, year y

KP(q, y) is the KP value for quarter q, year y

2.3.3 Notation for CYP series and PYP Series from monthly KP for periods after the last base
year for a given month

where:

PYP(m, y) is the PYP value for month m, year y

CYP(m, y) is the CYP value for month m, year y

KP(m, y) is the KP value for month m, year y

Page 7 of 22
Algebraically, for example, for a quarterly KP input series:

If LBY = the last base year, then for a chosen year greater than the last base year, for example y > LBY:

where

LBY is the last base year

PYP(q, y) is the PYP value for quarter q, year y

CYP(q, y) is the CYP value for quarter q, year y

KP(q, y) is the KP value for quarter q, year y

CP(B) is the annual CP value for the last base year

KP(B) is the annual KP value for the last base year

By definition of the last base year

And KP(LBY) is non-zero by definition, so

Therefore,

A similar approach is used for both the annual and the monthly KP input series cases.

3 . Deriving weights for aggregation

Chain-linking calculations are typically performed at the lowest level possible. Aggregates then need to be formed
from the lower-level calculations, which are achieved through the use of weights.

The weights of a chain volume measure are updated each year and are based on the fully available previous
year’s data. That is, the chain volume measure is not a fixed-weighted index with a single base year.

Chain Linked Weights, ONS (2014a) notes that: “Weights for 112 industry groups are calculated based on the
contribution of each industry to the overall economy. These are derived from gross value added (GVA) totals
produced by ONS in the Supply and Use Tables. GVA for an industry group is divided by total GVA across the
whole economy and then multiplied by 1000, to give a parts per thousand weight for that industry group. Weights
are created for all years that have been through at least 2 annual rounds of supply use balancing. The last year
we have weights for is also used as the reference year for the index (this is the convention for the UK but does
not have to be the case) … previous years weights are revised for the open period.”

See section 5 for discussion on the treatment, including the weighting, of series tails.

Page 8 of 22
4 . Chain-linking

For our outputs up to Blue Book 2016, the last base year is used as the reference year. From Blue Book 2017,
the chain-linking calculations within our systems will still process using the last base year as the reference year
but will do additional calculations to re-reference to a distinct separate reference year.

Within the UK National Accounts, we use Laspeyres chain-linking (Eurostat 2013b, Chapter 6). The defining
feature of Laspeyres is that in calculating growth from one period to another, the prices of the earlier period are
applied to both periods. The choice of an annually-chained Laspeyres index as the formula for computing annual
growth rates is straightforward and unambiguous. However, the specific linking methodology has to take into
account a number of factors including consistency with other series, monthly path and annual growth rates.

The methods used in our UK National Accounts are explained in later sections. However, the technical decisions
made for introducing annual chain-linking into our UK National Accounts appear in box 5 of Tuke and Reed
(2001), copied below:

Annual chain-linking will be carried out using Laspeyres indices and series will be referenced to the last
weights year to preserve additivity in the most recent data

Annual chain-linked quarterly data will be linked using an overlap on quarter four. Any resultant drift away
from the annual growth path will be removed by benchmarking to annual data. The decision was made to
use a quarterly overlap method rather than alternatives suggested by Eurostat because it best preserves
the quarterly growth path

Annual chain-linked monthly data will also be linked on quarter 4 and benchmarked to annual data. This
will minimise the impact of December to January effects and provides estimates which follow the quarterly
growth path

The level at which annual chain-linking will be implemented for GDP(O) will meet Eurostat requirements
and will be implemented at an even lower level where constant price data is consistently available and
modelling has shown that this chain-linking may significantly alter growth estimates

The monthly Index of Production will also be annual chainlinked to give consistency between monthly,
quarterly and annual published estimates

It should be noted that our Prices area derive price information using a different but appropriate methodology. For
further details Consumer Price Indices Technical Manual, ONS (2014b) and Ralph et. al. (2015).

4.1 Notation for annual series chain-linking for a given year

The annual chained volume measure (CVM) is defined

For low level series in may be possible that for a particular year CYP(y) = 0 and / or PYP(y) = 0. If for a particular
year CYP(y) = 0 and / or PYP(y) = 0, then set

Page 9 of 22
4.2 Notation for quarterly series chain-linking

A consistent approach to chain-linking has to be taken for the quarterly national accounts and short-term volume
indicators (such as the Index of Production, Output in the Construction Industry, the Index of Services, and the
Retail Sales Index), in order to ensure consistency between the different series. The quarterly path is best
calculated using the quarterly overlap technique (with quarter 4 (Oct to Dec) being the link period), as using an
annual overlap can introduce a jump between quarter 4 of one year and quarter 1 (Jan to Mar) of the next year
(for example, in Chapter 9 of Bloem et. al (2001) – in particular Example 9.4.a. on page 159.).

Applying the quarterly overlap to a monthly series ensures consistency with the quarterly series, while
maintaining a smooth monthly path. However, on its own, a quarterly overlap can lead to divergence between the
monthly (or quarterly) series and the annual series. Thus, the monthly (or quarterly) series is benchmarked to the
annual series in order to ensure consistency with the annual series and remove any divergence. This is done
automatically with the chain-linking function within our production systems. Consequently, quarterly overlap with
benchmarking, combines the principal advantages of both quarterly overlap (best intra-year path) and annual
overlap (consistency with annual data), and meets the requirements of key users of the data.

There is considerable discussion in the literature around aspects of the quarterly overlap method. For example,
Eurostat (2013b, Section 6.51) states:

“A key property of the one-quarter overlap method is that it preserves the quarter-on-quarter growth rate between
the fourth quarter of year y-1 and the first quarter of year y – unlike the annual overlap method. The ‘damage’
done to that growth rate by the annual overlap method is determined by the difference between the annual and
quarter overlap link factors. Conversely, this difference also means that the sum of the linked quarterly values in
year y-1 differ from the annual overlap-linked data by the ratio of the two link factors. Temporal consistency can
be achieved by benchmarking the quarterly chain-linked volume estimates derived using the one-quarter overlap
method to their annual counterparts. By using an optimal benchmarking procedure (see Chapter 5), the difference
between the two link factors for each year is spread over many quarters such that the amendments to the quarter-
on-quarter movements are minimized.”

Section 9.41, page 158, Bloem et. al (2001), states:

“To conclude, there are no established standards with respect to techniques for annually chain-linking of QNA
data, but chain-linking using the one-quarter overlap technique, combined with benchmarking to remove any
resulting discrepancies between the quarterly and annual data, gives the best result. In many circumstances,
however, the annual overlap technique may give similar results. The over-the-year technique should be avoided.”

As per standard chain-linking approaches, both the current year’s price (CYP) and price year’s price (PYP)
quarterly series are additive. The annual value may be obtained from summing the individual quarters.

Hence,

The Unconstrained CVM series is an intermediate series that is produced before being constrained by
benchmarking to the annual CVM series to produce the final CVM.

Where the ratio

Page 10 of 22
and where LBY = last base year.

In the standard situation where the annual series and quarter 4 (Oct to Dec) values for both the PYP and CYP
are never zero, for example in the case of high level aggregates, then the algebra may be simplified to

An annual version of the input series are then calculated, then these annual series are then chain linked using the
method in section 4.1.

The quarterly unconstrained CVM is then benchmarked to the annual chain linked series to ensure consistency.

An important user requirement is that the quarterly system of accounts should provide high quality measures of
growth, with no discontinuities, and it should allow for growth to be estimated over varying period lengths. The
Cholette-Dagum method (Cholette and Dagum, 1994), a special case of Denton, is now the default method used
for benchmarking within our National Accounts. The constrained chained volume measure is the benchmarked
series for time periods up to and including the last base year. The constrained chained volume measure is the
unconstrained chained volume measure for time periods after the last base year.

4.3 Notation for monthly series chain-linking

The UK short-term volume indicators (for example the Index of Production, Output in the Construction Industry,
the Index of Services, and the Retail Sales Index) are fully integrated with the quarterly and annual national
accounts; that is, monthly estimates are constrained to sum to annual estimates, both in terms of current prices
and CVMs. It is therefore an important requirement for the UK that the following basic accounting conventions are
preserved within a chain-linked system.

As per standard practice, both the CYP and PYP series are additive. A quarterly value may be obtained from
summing the associated monthly values. For example,

In this situation the unconstrained CVM is defined as:

Where the ratio

Page 11 of 22
and where LBY = last base year.

An annual version of the input series can be calculated, where these annual series are then chain linked using
the method in 4.1. The quarterly unconstrained CVM is then benchmark to the annual chain linked series. Similar
to the quarterly benchmarking described in 4.2.

5 . Tail of series up to Blue Book 2016

Up to and including the last base year we have full years of data and weights, the calculations for the estimates
follow the forms as described in the earlier sections. After the last base year a fixed base volume calculation is
used for the chained volume measure (CVM) series – this is referred to as the “tail” or the “Fixed Base Tail”. In
effect two methods are joined together in the last base year.

Our treatment of the tail is covered in Documentation on Current Methods used for National Accounts, ONS
(2008, Section 2.8.2) which notes:

“The same reference year is used for all National Accounts aggregates. When annual chain-linking was
introduced to the UK National Accounts, the decision was made, after consultation with external users, to make
the reference year the same as the last year from which weights have been used in the aggregate time series
(and adopt this reference year with lower level time series, regardless of the last year of weights which had been
used in their construction). This allows for additivity in the latest time periods, i.e. component levels add to
aggregate levels for £m series, and for index series the aggregates can be calculated by weighting together the
separate components. This additivity is found in the reference year onwards for annual levels and in the
reference year plus one onwards for quarterly and monthly levels. In contrast with the use of CVMs with output or
consumption valued at the prices in the reference year, when volume changes are measured relative to a fixed
base year, the volume series re-valued at the prices in the base year are termed constant price (KP) estimates.”

The calculation of the tail of the series is not covered explicitly in the international manuals. However, Eurostat
(2013b, Sections 6.85 and 6.86) states:

“The requirements to derive estimates for a quarter in the prices of the previous year are volume indices from the
base year to the quarter and price (or current price) data for the base year – see formula 6.14. These
requirements are always met for the final expenditure components of GDP(E), but the price data may not be
available for the gross value added by industry components of GDP(P). This is likely to be the case if quarterly
current price estimates of gross value added are not derived and the quarterly volume estimates of gross value
added are derived, in the main, by extrapolation using volume indicators of output. In such cases, the timing of
the introduction of a new base year may need to coincide with the introduction of new annual estimates, including
first estimates of gross value added for the new base year.

“The timing and consequences of this approach is best explained using an example. Suppose in country A new
annual estimates and a new base year are introduced with the release of data for the second quarter each year.
Further suppose the current year is year y and the new annual estimates are in respect of year y-2, then country
A may choose to make the year y-2 the new base year. This means that (prior to linking) the volume estimates
for the four quarters in year y-1 and the first two quarters of year y are now derived in the average prices of year
y-2. By contrast, in the preceding release (that is that for the first quarter of year y) the volume estimates for the
four quarters of year y-1 and the first quarter of year y were derived in the average prices of year y-3. The change
of base year, from year y-3 to year y-2, is likely to change the growth rates of the five quarters concerned. For the
majority of national accounts statistics, price and volume relativities do not change very much between one year
and the next, and so in most cases the revisions to growth rates are small.”

Page 12 of 22
Within our context, for Blue Books up to Blue Book 2016, the last base year is the same as the reference year.
For example, for Blue Book 2016, the last base year is 2013. In which case the

CVM series = CYP series = PYP series

for periods after 2013. This holds for monthly, quarterly and annual series. As noted previously, the previous year’
s prices (PYP) series is defined the same as current year’s prices (CYP) series after the last base year and so is
thus not truly “previous year’s prices” despite the name.

The tail is a constant price series as it is based on fixed-price weights which are the same as those used for the
last base year, 2013.

For Blue Book 2017 onwards, section 6 describes the calculations required in relation to the tail estimates.

6 . Re-referencing and the tail calculation from Blue Book

2017

6.1 Background on tail calculation

Drew et. al (2016) describes this issue in more detail and provides an explanation of the change in approach to
UK outputs.

ESA 2010 (Eurostat, 2013c) states that chained volume measures (CVMs) should be formed the prices of the
previous year. For the UK, the chain-linking function has two methods available, the correct annual chain-linking
method up until the last base year and a fixed base method for the tail, which was defined when annual chain-
linking was first introduced in 2003.

The issue of previous year’s prices was not included in ESA 1995 (Eurostat, 1999) and was covered instead by
the Price and Volume Handbook in 2001 (Eurostat, 2001). This handbook mandated previous years for annual
data only and became a legislative requirement from 2006 for the UK. The Quarterly National Accounts
Handbook is the only Eurostat handbook to mention the issue of the tail, and sets out the option of having a tail.
ESA 2010 sets out in 10.20:

“Therefore, the calculation of the volume is made only for two successive years, that is the volume is calculated
at the prices of the previous year.”

There is no discussion of the practical issues surrounding this implementation and has been interpreted as
applying to all years. Eurostat (2013a) makes no changes in this area to the 2001 version.

6.2 Calculations for the Reference Year from Blue Book 2017 onwards

From Blue Book 2017 onwards the reference year and last base year will be able to be distinct and different.

In practice, the reference year is a presentational reference point where the annual CP = annual CVM.

From Blue Book 2017 the reference year is the first year balanced in the supply and use tables (SUT) framework.
For example, for Blue Book 2017 the Reference Year will be 2015 which is the initial balanced year in SUT (t-2)
and 2014 is the re-balanced year (t-3).

Page 13 of 22
Note that many other countries keep their reference year fixed for many years. Some countries update in line with
large benchmarking exercises every five years.

6.3 Changes to UK production systems

UK production systems will continue to process constrained CVMs as previously used (and described in Sections
2 to 6 in this note) and will re-reference the constrained CVMs to desired reference year.

6.3.1 Calculations for re-referencing annual series

The re-referencing is achieved by the calculation:

where

ref is the reference year

CVMRE(y) is the annual referenced CVM value for year y

CVM(ref) is the annual CVM value for the reference year

CP(ref) is the annual CP value for the reference year

CVM(y) is the annual CVM value for year y

6.3.2 Calculations for re-referencing quarterly series

The re-referencing is achieved by the calculation:

where

ref is the reference year

CVMRE(q, y) is the referenced CVM value for quarter q, year y

CVM(ref) is the annual CVM value for the reference year

CP(ref) is the annual CP value for the reference year

CVM(q, y) is the CVM value for quarter q, year y

6.3.3 Calculations for re-referencing monthly series

where

Page 14 of 22
 ref is the reference year

CVMRE(m, y) is the referenced CVM value for month m, year y

CVM(ref) is the annual CVM value for the reference year

CP(ref) is the annual CP value for the reference year

CVM(m, y) is the CVM value for month m, year y

6.4 Approach to generate previous year price’s that are consistent with re-
referenced constrained CVM series and with the desired tail properties

Due to the way our production systems are configured the most efficient way to generate the required version of
the previous year price’s series is to perform additional calculations to meet the required properties.

The notation CVMRE gives the constrained CVM values that have been referenced to a reference year that is
distinct from the last base year.

From Blue Book 2017 onwards the CVMRE series are the published chain-linked estimates. The corresponding
PYP series published, which relate to the CVMRE series, are described below.

6.4.1 Calculations for previous year’s price annual series aligned to CVMRE

The previous year’s price based on this reference year is calculated as:

where

y-1 is the year previous to year y

PYP*(y) is the annual PYP value for year y consistent with the referenced values

CVMRE(y) is the referenced CVM value for year y

CP(y-1) is the annual CP value for year y-1

CVMRE (y-1) is the referenced CVM value for year y-1

6.4.2 Calculations for previous year’s price quarterly series aligned to CVMRE

The previous year’s price based on this reference year is calculated as:

where

Page 15 of 22
 y-1 is the year previous to year y

PYP*(q, y) is the PYP value for quarter q, year y consistent with the referenced values

CVMRE(q, y) is the referenced CVM value for quarter q, year y

CP(y-1) is the annual CP value for year y-1

CVMRE (y-1) is the referenced CVM value for year y-1

6.4.3 Calculations for previous year’s price monthly series aligned to CVMRE

The previous year’s price based on this reference year is calculated as:

where

y-1 is the year previous to year y

PYP*(q, m) is the PYP value for month m, year y consistent with the referenced values

CVMRE(q, m) is the referenced CVM value for month m, year y

CP(y-1) is the annual CP value for year y-1

CVMRE (y-1) is the referenced CVM value for year y-1

6.5 Property of the previous year price’s for the year after the reference year

If y is the year after the reference year, that is y = ref +1, then

Where

ref is the reference year

PYP*(q, y) is the PYP value for quarter q, year y consistent with the referenced values

CVMRE(q, y) is the referenced CVM value for quarter q, year y

CVM(ref) is the annual CVM value for the reference year

CP(ref) is the annual CP value for the reference year

CVM(q, y) is the CVM value for quarter q, year y

That is, in the year after the reference year the previous year’s price value is the same as re-referenced
constrained volume measure value.

Page 16 of 22
7 . National Accounts Identities and consistency checks

Published National Accounts estimates should always adhere to standard identities. The following sections
outline identities relevant to chained volume measure (CVM) and current price (CP) estimates.

7.1 Chain-linking one CYP series with a corresponding PYP series where the
last base year is the reference year. Where in the tail the PYP series = CVM
series = CYP series. That is, the method used up to Blue Book 2016

Conceptually, the CP series and current year’s prices (CYP) series are not the same

Annual CP series = annual CVM in the last base year

Annual CVM = annual CYP in the last base year and subsequent periods

In the last base year the monthly CVM components sum to the annual CVM value

In the last base year the quarterly CVM components sum to the annual CVM value

In the last base year or for periods before the last base year: monthly CVM components do not necessarily
add to the quarterly CVM value

After the last base year CVM series are additive, that is, monthly may be summed to quarterly or annual,
quarterly may be summed to annual

CVM series = previous year’s prices (PYP) series = CYP series post the last base year. This holds for
monthly, quarterly and annual series

CYP and PYP individual time series are additive over time, that is, monthly may be summed to quarterly or
annual, quarterly may be summed to annual

PYP series actually use prices of the last base year for all years after the last base year (so not "previous
year’s prices" in the tail despite the name). That is, the tail is actually the CYP series

Annual CP = annual CYP up to and including the last base year

7.2 Aggregation of series where the last base year is the reference year. Where
in the tail the PYP series = CVM series = CYP series. That is, the method used
up to Blue Book 2016

Page 17 of 22
 CVMs are additive post the last base year, that is, after the last base year lower level component CVM
time series, can be weighted and summed to create aggregate a higher level CVM time series

In the last base year or for periods before the last base year: lower level component CVM series are not
additive, that is, lower level component CVM time series, cannot be weighted and summed to create
aggregate higher level CVM time series

PYP series are always additive, that is, lower level component PYP time series, can be weighted and
summed to create aggregate higher level PYP time series

CYP series are always additive, that is, lower level component CYP time series, can be weighted and
summed to create aggregate higher level CYP time series

Annual SA totals are forced to equal annual NSA totals, our convention, in line with international best
practice (for example the QNA manual) is that annual totals for seasonally adjusted series are forced by
benchmarking to equal annual totals of the corresponding non-seasonally adjusted series

7.3 Chain-linking one CYP series with a corresponding PYP series where the
last base year is distinct from the reference year. That is, the method used
from Blue Book 2017

When the last base year is distinct from the reference year, CVMRE are constrained CVM values that have been
referenced to a reference year that is distinct from and later than the last base year.

Conceptually, the CP series and CYP series are not the same

Annual CP series = annual CVMRE in the reference year

Annual CP = annual CYP in the reference year

In the last base year and reference year CVMRE series are additive, that is, monthly may be summed to
quarterly CVMRE value or annual CVMRE value, quarterly CVMRE may be summed to annual CVMRE
value

For periods that are not in the last base year or reference year: monthly CVMRE components are not
additive, so will not necessarily sum to the quarterly CVMRE or the annual CVMRE value

For periods that are not in the last base year or reference year: quarterly CVMRE components are not
additive, so will not necessarily sum to the annual CVMRE value

CVM series = PYP series in the year after the reference year. (PYP as in 6.5.)

CYP and PYP individual time series are additive over time. That is, monthly may be summed to quarterly
or annual, quarterly may be summed to annual

7.4 Aggregation of series where the last base year is distinct from the
reference year. That is, the method used from Blue Book 2017

When the last base year is distinct from the reference year, CVMRE are constrained CVM values that have been
referenced to a reference year that is distinct from the last base year.

Page 18 of 22
 CVMRE are additive for the reference year and the last base year, that is, in the reference year or the last
base year level lower component CVMRE time series, can be weighted and summed to create aggregate a
higher level CVMRE time series

For periods other than the reference year or the last base year: lower-level component CVMRE series are
not necessarily additive, that is, lower-level component CVMRE time series cannot necessarily be weighted
and summed to create aggregate higher level CVMRE time series

PYP series are always additive, that is lower level component PYP time series, can be weighted and
summed to create aggregate higher level PYP time series

CYP series are always additive, that is lower level component CYP time series, can be weighted and
summed to create aggregate higher level CYP time series

Annual SA = annual NSA totals, that is, annual totals for seasonally adjusted series are forced to equal
annual totals of the corresponding non-seasonally adjusted series.

8 . Summary

The same chain-linking methodology is used consistently throughout the UK national accounts. To chain-link
quarterly (or monthly) series the "quarterly overlap" technique is used with benchmarking to the annual chain-
linked series. It is internationally recognised as the method which gives the best results, with a smooth quarter 4
to quarter 1 transition and annual growth rates which are consistent with the annual series.

This is also recommended as part of the 2008 edition of the United Nations System of National Accounts (UN,
2008) states in paragraph 15.50: “…, chaining using the one-quarter overlap technique with benchmarking to
remove any resulting discrepancies between the quarterly and annual data gives the best result.”

9 . Acknowledgements

The authors would like to thank all our colleagues, including Rob Bucknall, Pete Lee, and David Matthews, for
their feedback and comments in preparing this article. Of course, any errors and omissions are our responsibility.

10 . Appendix ONS National Accounts notation and

terminology
Term Description

Page 19 of 22
Base Year /Period The year /period in which volume series are expressed using the
value of the average prices.
CP Current price.
Estimates valued in the prices of the period when the activity
occurred. Also referred to as “Nominal Prices”.
CVM Chained volume measure.
The result of joining together two indices that overlap in one period
by rescaling one of them to make its value equal to that of the other
in the same period, thus combining them into single time series.
Also referred to as “Chain-Linked Values”.
CVMRE Chained volume measure that has been re-referenced to a
reference year distinct from the last base year.
A volume measure obtained by chain linking.
CYP Current year's prices.
The average price in the year in which the activity took place.
Deflator A price index.
By applying the deflator to a current price series the associated
volume measure series may be calculated.
Fixed Base Index An index which uses fixed weights from the defined base period for
all component index numbers.
KP Constant Price Series (KP).
Term referring to expressing values in terms of a base period. A
volume also known as constant prices or real growth. Calculated
directly or indirectly by dividing current price value by the deflator
(price index). In effect you are holding the prices constant in the
base period or removing the effect of price change.
Last base year This is the last year in a Chained Volume Measure Series which is
used as a base year.
In the last base year annual CP value = annual KP value.
NSA Non-seasonally adjusted.
PYP Previous year's prices.
The average price in the year preceding the period in which the
activity took place. The series derived from multiplying the volume
series in the current period, by the average price of the previous
year. This process is also known as unchaining. Previous Years
Prices are additive and are the method by which volume series can
be aggregated together.
PYP* Previous year's prices that are consistent with the re-referenced
CVM and the tail are genuine previous year's prices.
Reference period The reference period is the period for which an index series is set
equal to 100 or the period for which a volume index series may be
set equal to the current price value in order to express the index
series in terms of currency units. In the reference year, the implied
deflator is equal to 1, and series are additive.

Page 20 of 22
SA Seasonally Adjusted.
Seasonal adjustment aids interpretation by removing effects
associated with the time of the year or the arrangement of the
calendar, which could obscure movements of interest.
Volume Index At the elementary level (that is, detailed disaggregated level) a
volume index is most commonly presented as a weighted average
of the proportionate changes in the quantities of a specified set of
goods or services between two periods of time. In a volume
measure, the estimates for all periods are in the same price.
VM Volume measure.

11 . References and further reading

Bloem, A.M, Dippelsman, R.J., and Maehle N.O. (2001) Quarterly National Accounts Manual—Concepts , Data
Sources, and Compilation, International Monetary Fund 2001.

Butler, R. (2012) Chain-linking of UK short term volume series, Office for National Statistics, 2012 Internal Paper.

Cholette, P.A. and Dagum, E.B. (1994). Benchmarking time series with autocorrelated survey errors, International
Statistical Review 62, 3: 365-377.

Daniel, E. (2014) An introduction to reconciled estimates of GDP , ONS, March 2014.

Drew, S., Hughes, M., and Denley H., (2016) Changes to Chain Volume Measure calculations for UK
Publications and International Transmissions from Blue Book 2017, Office for National Statistics, October 2016.

Eurostat, (2013a), Handbook on Prices and Volume Measures in National Accounts .

Eurostat, (2013b), Handbook on Quarterly National Accounts .

Eurostat, (2013c), European system of accounts - ESA 2010 .

Eurostat, (1999), European system of accounts ESA 1995 .

Eurostat, (2001), Handbook on price and volume measures in national accounts .

Lee, P. (2011). United Kingdom National Accounts – a short guide , Office for National Statistics, August 2011.

Office for National Statistics (2008), Documentation on Current Methods used for National Accounts , April 2008,
UK Centre for the Measurement of Government Activity.

Office for National Statistics (2014a), Chain Linked Weights , January 2014.

Office for National Statistics, (2014b), Consumer Price Indices Technical Manual .

Page 21 of 22
Office for National Statistics (2015a), Economic Review , Office of the Chief Economic Adviser. Office for National
Statistics, 05 August 2015.

Office for National Statistics (2015b), Economic Review, Illustrative Example of a CVM Calculation in National
Accounts.

Mahajan, S. (2006), Development, compilation and use of Input-Output Supply and Use tables , Office for National
Statistics, Economic Trends 634, September 2006.

Ralph, J. O’Neill R., and Winton J. (2015), A Practical Introduction to Index Numbers, ISBN: 978-1-118-97781-1

Robjohns J., Methodological Note: Annual chain-linking , ONS, Economic Trends, No 630, 2006.

Scheiblecker, M., 2010, Chain-linking in Austrian quarterly national accounts and the business cycle , OECD
Journal: Journal of Business Cycle Measurement and Analysis, No.: 5, Volume: 2010, Issue: 1 Pages 1–12.

Tuke A. and Reed, G. (2001), The effects of annual chain-linking on the output measure , Office for National
Statistics, Economic Trends.

United Nations, 2008, System of National Accounts – SNA 2008 .

Page 22 of 22